X-Git-Url: http://git.nguyen.vg/gitweb/?a=blobdiff_plain;f=main.ml;h=4fcc0af9e2ed5631c5b2886e691ffec8ab270b94;hb=70ff0bfc463882ecf233f1b1a7ac4a8007fa4cc2;hp=7cc631b0f1e9c6b0191eeeb7728145f438f11007;hpb=4a66518948bf6356b5cb72ba30b4d48a8c9e833a;p=SXSI%2Fxpathcomp.git diff --git a/main.ml b/main.ml index 7cc631b..4fcc0af 100644 --- a/main.ml +++ b/main.ml @@ -4,237 +4,119 @@ (* Copyright NICTA 2008 *) (* Distributed under the terms of the LGPL (see LICENCE) *) (******************************************************************************) -INCLUDE "debug.ml" open Ata - let l = ref [] ;; let time f x = let t1 = Unix.gettimeofday () in let r = f x in let t2 = Unix.gettimeofday () in - let t = (1000. *.(t2 -. t1)) in + let t = (1000. *. (t2 -. t1)) in l:= t::!l; Printf.eprintf " %fms\n%!" t ; r ;; let total_time () = List.fold_left (+.) 0. !l;; +let enabled_gc = Gc.get() +let disabled_gc = { Gc.get() with + Gc.max_overhead = 1000000; + Gc.space_overhead = 100 } -let poa = TagSet.add Tag.pcdata (TagSet.singleton Tag.attribute) - -let rec fill_hashtag t = - if Tree.Binary.is_node t then - begin - let tag = Tree.Binary.tag t in - let a = - if TagSet.mem tag poa - then 0 - else - fill_hashtag (Tree.Binary.first_child t) - in - let b = fill_hashtag (Tree.Binary.next_sibling t) - in a+b+1 - end - else 0 - - -let test_slashslash tree k = - let test = - match k with "*" -> TagSet.remove (Tag.tag "") TagSet.star - | s -> TagSet.singleton (Tag.tag k) - in - let attorstring = TagSet.cup TagSet.pcdata TagSet.attribute in - let rec aux t acc = - if Tree.Binary.is_node t - then - let tag = Tree.Binary.tag t in - let l = Tree.Binary.first_child t - and r = Tree.Binary.next_sibling t - in - let acc = - if TagSet.mem tag test - then - TS.append t acc - else - acc - in - let rl = if TagSet.mem tag attorstring then acc else aux l acc - in aux r rl - else - acc - in - let _ = Printf.eprintf "Testing optimal //%s ... " k in - let r = time (aux tree ) TS.empty in - Printf.eprintf "Result set is %i nodes\n%!" (TS.length r) - - -let test_jump tree k = - let ttag = Tag.tag k in - - let rec loop acc tree = - if Tree.Binary.is_node tree - then - let acc = TS.cons tree acc in - loop acc (Tree.Binary.tagged_foll tree ttag) - else - acc - - in - let _ = Printf.eprintf "Testing jumping for tag %s ... " k in - let r = time (loop TS.empty ) (Tree.Binary.tagged_next tree ttag) in - Printf.eprintf "Result set is %i nodes\n%!" (TS.length r) - - - -let test_traversal tree k = - let ttag = Tag.tag k in - let iid t = if Tree.Binary.is_node t then Tree.Binary.id t else -1 in - let rec aux t = - if Tree.Binary.is_node t - then - let tag = Tree.Binary.tag t in - let l = Tree.Binary.first_child t - and r = Tree.Binary.next_sibling t - in - let _ = Printf.eprintf "Tree with id %i and tag=%s, tagged_desc %s is %i tagged_foll is %i\n%!" - (Tree.Binary.id t) (Tag.to_string tag) (k) - (iid (Tree.Binary.tagged_desc t ttag)) - (iid (Tree.Binary.tagged_foll t ttag)) - in - aux l; - aux r; - - else - () - in - aux tree - - - -let test_count_subtree tree k = - let ttag = Tag.tag k in - let _ = Printf.eprintf "Counting subtrees with tag %s ... %!" k in - let r = time(Tree.Binary.subtree_tags tree) ttag in - Printf.eprintf "%i nodes \n%!" r - - -let test_contains tree s = - let _ = Printf.eprintf "Fetching DocIds containing %s ... %!" s in - time (fun s -> let r = Tree.Binary.contains tree s in - Tree.Binary.DocIdSet.iter - (fun t -> output_string stderr - (Tree.Binary.get_string tree t); - output_char stderr '\n') r ) s - - -let test_count_contains tree s = - let _ = Printf.eprintf "Counting DocIds containing %s ... %!" s in - let r = time (Tree.Binary.count_contains tree) s in - Printf.eprintf "%i documents ids\n%!" (r) - -let test_contains_old tree s = - let _ = Printf.eprintf "Fetching (old) DocIds containing %s ... %!" s in - let r = time (Tree.Binary.contains_old tree) s in - Printf.eprintf "%i documents ids\n%!" (Tree.Binary.DocIdSet.cardinal r) - -let test_contains_iter tree s = - let _ = Printf.eprintf "Fetching (old) DocIds containing %s ... %!" s in - let r = time (Tree.Binary.contains_iter tree) s in - Printf.eprintf "%i documents ids\n%!" (Tree.Binary.DocIdSet.cardinal r) - -module Stack = -struct - type t = { mutable table: Tree.Binary.t array; - mutable top : int } - - let empty = { table = Array.make 0 (Obj.magic 0); - top = 0 } - let cons e s = - let ls = Array.length s.table in - if ls > s.top - then - begin - s.table.(s.top) <- e; - s.top <- s.top + 1; - s - end - else - - let a = Array.make (ls * 2 + 1) (Tree.Binary.root e) - in - Array.blit s.table 0 a 0 ls; - s.table <- a; - s.table.(s.top) <- e; - s.top <- s.top + 1; - s - -end - -let test_fast tree = - let rec aux t acc = - if Tree.Binary.is_node t - then - aux (Tree.Binary.right t)( aux (Tree.Binary.left t) (Stack.cons t acc)) - else acc - in let _ = Printf.eprintf "Fast traversal ...%!" in - time (aux tree) Stack.empty - -let test_cps tree = - let rec aux t acc cont = - if Tree.Binary.is_node t - then aux (Tree.Binary.left t) (Stack.cons t acc) ((Tree.Binary.right t)::cont) - else match cont with - | [] -> acc - | p::r -> aux p acc r - in - let _ = Printf.eprintf "CPS traversal ...%!" in - time (aux tree Stack.empty) [] - - -let main v query output = - let _ = Tag.init (Tree.Binary.tag_pool v) in +let main v query_string output = + let _ = Tag.init (Tree.tag_pool v) in Printf.eprintf "Parsing query : "; let query = try time - XPath.Parser.parse_string query + XPath.Parser.parse_string query_string with Ulexer.Loc.Exc_located ((x,y),e) -> Printf.eprintf "character %i-%i %s\n" x y (Printexc.to_string e);exit 1 in XPath.Ast.print Format.err_formatter query; Format.fprintf Format.err_formatter "\n%!"; -(* Printf.eprintf "Dummy iteration : "; - time (fill_hashtag) v; - Printf.eprintf "Dummy iteration (tag access cached) : "; - time (fill_hashtag) v; -*) Printf.eprintf "Compiling query : "; - let auto,_ = time XPath.Compile.compile query in - Printf.eprintf "Execution time %s : " (if !Options.count_only then "(counting only)" else ""); + let auto,ltags,contains = time (XPath.Compile.compile ~querystring:query_string) query in + let _ = Ata.dump Format.err_formatter auto in + let _ = Printf.eprintf "%!" in + let jump_to = + match contains with + None -> (max_int,`NOTHING) + | Some s -> + let r = Tree.count v s + in + Printf.eprintf "%i documents in the TextCollection\n" (Tree.text_size v); + Printf.eprintf "Global count is %i, using " r; + if r < !Options.tc_threshold then begin + Printf.eprintf "TextCollection contains\nCalling global contains : "; + time (Tree.init_contains v) s; + end + else begin + Printf.eprintf "Naive contains\nCalling global contains : "; + time (Tree.init_naive_contains v) s + end;(r,`CONTAINS(s)) + in + let test_list = + if (!Options.backward) then begin + Printf.eprintf "Finding min occurences : "; + time + ( List.fold_left (fun ((min_occ,kind)as acc) (tag,_) -> + let numtags = Tree.subtree_tags v tag Tree.root in + if ((numtags < min_occ) && numtags >= 2) + then (numtags,`TAG(tag)) + else acc) jump_to) ltags + end + else (max_int,`NOTHING) + in + let _ = if (snd test_list) != `NOTHING then + let occ,s1,s2 = match test_list with + | (x,`TAG (tag)) -> (x, "tag", (Tag.to_string tag)) + | (x,`CONTAINS(s)) -> (x, "contains()", s) + | _ -> assert false + in + Printf.eprintf "Will jump to %s %s occuring %i time\n%!" s1 s2 occ + in + Printf.eprintf "Execution time %s : " + (if !Options.count_only then "(counting only)" else if !Options.backward then "(bottomup)" else ""); begin - if !Options.count_only then - let result = time (BottomUpNew.run_count auto) v - in - Printf.eprintf "Number of nodes in the result set : %i\n" result - else - let result = time (BottomUpNew.run auto) v in - Printf.eprintf "Number of nodes in the result set : %i\n" (TS.length result); - begin - match output with - | None -> () - | Some f -> - Printf.eprintf "Serializing results : "; - time( fun () -> - let oc = open_out f in - output_string oc "\n"; - TS.iter (fun t -> output_string oc "----------\n"; - Tree.Binary.print_xml_fast oc t; - output_char oc '\n') result) (); - end; + let _ = Gc.full_major();Gc.compact() in + (* let _ = Gc.set (disabled_gc) in *) + if !Options.backward && ((snd test_list) != `NOTHING )then + + let r = time (bottom_up_count auto v )(snd test_list) in + let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r + in () + else + let _ = + if !Options.backward then Printf.eprintf "WARNING: couldn't find a jumping point, running top-down\n" + in + if !Options.count_only then + let r = time ( top_down_count auto ) v in + let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r + in () + else + let result = time (top_down auto) v in + let rcount = IdSet.length result in + Printf.eprintf "Number of nodes in the result set : %i\n" rcount; + Printf.eprintf "\n%!"; + begin + match output with + | None -> () + | Some f -> + Printf.eprintf "Serializing results : "; + time( fun () -> + let oc = open_out f in + output_string oc "\n"; + IdSet.iter (fun t -> + Tree.print_xml_fast oc v t; + output_char oc '\n'; + output_string oc "----------\n"; + ) result) (); + end; end; - Printf.eprintf "Total running time : %fms\n%!" (total_time()) + let _ = Gc.set enabled_gc in + Printf.eprintf "Total running time : %fms\n%!" (total_time()) ;; - + Options.parse_cmdline();; let v = @@ -242,28 +124,21 @@ let v = then begin Printf.eprintf "Loading from file : "; - time (Tree.Binary.load ~sample:!Options.sample_factor ) - (Filename.chop_suffix !Options.input_file ".srx"); - end + time (Tree.load ~sample:!Options.sample_factor ) + (Filename.chop_suffix !Options.input_file ".srx"); + end else let v = - time (fun () -> let v = Tree.Binary.parse_xml_uri !Options.input_file; + time (fun () -> let v = Tree.parse_xml_uri !Options.input_file; in Printf.eprintf "Parsing document : %!";v ) () in if !Options.save_file <> "" then begin Printf.eprintf "Writing file to disk : "; - time (Tree.Binary.save v) !Options.save_file; + time (Tree.save v) !Options.save_file; end; v in main v !Options.query !Options.output_file;; -IFDEF DEBUG -THEN -Printf.eprintf "\n=================================================\nDEBUGGING\n%!"; - -Tree.DEBUGTREE.print_stats Format.err_formatter;; -Gc.full_major() -ENDIF